Effect of Subprime Crisis on U.S. Stock Market Return and Volatility

Transcription

1 Global Economy and Finance Journal Vol 4. No. 1. March 2011 Pp Effec of Subprime Crisis on U.S. Sock Marke Reurn and Volailiy Choy Yoke Chong* The U.S. Subprime credi crisis began wih massive defauls by Subprime borrowers in morgage markes. The crisis creaed cerain degree of impac on financial markes as shown by Longsaff (2010) on he conagion across financial markes from he credi crisis. Hence, his paper aemped o sudy abou he behavior of U.S. sock marke in erm of sock reurn and volailiy along wih he Subprime crisis. Daily daa of S&P 100 sock indexes for he period of May 2006 o December 2009 are sudied using basic GARCH model. The daa were hen divided ino hree differen sub-periods o allow he behavior of sock marke in differen sub-periods o be invesigaed. The following sub-periods are idenified: early sage of Subprime crisis, recession of U.S. and period afer he bankrupcy of Lehman Brohers. This paper revealed ha he bankrupcy of Lehman Brohers following he Subprime crisis urned ou o have bigger impac on sock marke volailiy bu no on he sock reurns in general, hough he impac is ransiory. JEL Codes: G01, G10 and G15 1. Inroducion Basically, he U.S. Subprime credi crisis sared wih huge defauls by Subprime borrowers in he morgage markes. The credi crisis led o massive losses or even bankrupcy among financial insiuions and companies which hold large porfolio wih morgage-backed securiies. This is consisen wih he evidence provided by Longsaff (2010) of he conagion across markes from he credi crisis. Therefore, researchers gain ineres o sudy abou he U.S. Subprime credi crisis in heir sock marke analysis. Besides, various researches were conduced o sudy on sock marke wih large shocks due o global even or crisis such as he crash of 1987 (Choudhry 1996, Law 2006), he Asian crisis (Chakrabari & Roll 2002, Holden e al 2005, Law 2006, Leeves 2007 and Karunanayake e al 2010), he Sepember 11 h erroris aacks (Charles & Darne 2006 and Nikkinen e al 2008). From ha respec, his aricle inends o sudy abou he behavior of sock marke in erm of he sock reurn and volailiy wih he onse of Subprime crisis. Iniial sign of he Subprime credi crisis was deeced on he real esae marke when a drop of 3.3 percen on median price of U.S. home from he fourh quarer of 2005 and he real esae gains came o an abrup hal in he firs quarer of 2006 as repored by Chrisie via CNNMoney.com daed May 16, However, he credi crunch had slow conagion effecs across oher financial markes since he official recession was repored by The Naional Bureau of Economic Research (NBER) afer 2 years ime a *Choy Yoke Chong, Faculy of Managemen, Mulimedia Universiy, Malaysia. cychong@mmu.edu.my

2 December 2007 (Isidore 2008). Hence, we can furher sudy he impac of Subprime crisis on sock marke behavior during recession. In addiion, he bankrupcy of Lehman Brohers, he fourh larges U.S. invesmen bank a Sepember 15, 2008 urned ou o be anoher big crash caused by he morgage crisis leading a deep fall on sock marke price (refer Figure 1). Thus, i would be ineresing o know he exen o which he Subprime crisis has shape he sock marke behavior in erm of reurns and volailiy in differen sages. As a resul, his paper firsly would like o sudy wheher here is volailiy clusering deeced in differen sages of Subprime crisis in U.S. sock markes. Nex, he persisence of shocks o he volailiy of reurns is invesigaed if any. Las bu no leas, comparison on he behavior of U.S. sock marke in erms of reurn and volailiy is carried ou across differen sages of Subprime crisis or sub-periods. In his conex, his paper is organized as follows. Secion 2 summarized some previous research oucomes done by he scholars relaing sock marke analysis wih sock marke crashes or global evens. Nex, he daa examined and saisical analysis for answering he research quesions in his sudy are described in secion 3. The following secion explained he resuls and findings for he sudy. Lasly, he paper ends wih some conclusion remarks and limiaion of sudy. 2. Lieraure Review 2.1 Sock Marke Analysis and Marke Crash or Crisis Choudhry (1996) had developed a research on sock marke volailiy wih he crash of Six emerging markes were sudied using a GARCH-M model o invesigae abou he volailiy, presence of ime-varying risk premium and he persisence of shocks o he volailiy before and afer he crash of Overall, significan volailiy clusering was found in sock reurns bu he shocks were no explosive. Some markes like Argenina, Greece and India demonsraed an increase of ARCH (Auoregressive Condiional Heeroskedasiciy) effecs afer he crash while he res of he emerging markes indicaed significan change in ARCH effecs (has/has no) before and afer crash. Indeed, he findings only managed o find some evidence of changes in persisence of shocks o volailiy for some emerging markes before and afer he crash. On he oher hand, he resuls mosly failed o prove any significan influence of volailiy on reurns (risk premia) excep some significan inverse relaionship was found for Argenina and India in poscrash. Besides, Asian crisis also generaed grea ineres for he researcher o sudy abou he performance of sock marke. Chakrabari and Roll (2002) did a comparison beween he Eas Asian sock markes (reaed as diseased paien of he 1997 Asian crisis) and European markes (reaed as healhy paien). Their resuls repored an increase of reurn volailiy and co-movemen for boh regions wih he onse of he crisis. No surprisingly, he magniude of increases in boh measuremens is 4 imes larger in Eas Asia. In addiion, diversificaion poenial or invesmen opporuniies were found o decline grealy a Eas Asia during he crisis. Holden e al (2005) on he oher hand sudied he calendar effecs due o he dayof-week, monh-of-year, days before or afer holidays and wihin monh effecs on daily sock marke reurns of Thailand. Paricular aenion of hese effecs was 103

3 given prior o, during and afer he Asian crisis. The resuls indicaed many of esed calendar effecs were small and no significan bu hese effecs can significanly improve he forecas performance. The behavior of sock reurns in erms of he ARCH effecs, serial correlaion problem and funcional form urned ou o be differen before, during and afer Asian crisis. Law (2006) also documened a prolonged impac on sock marke volailiy in Malaysia due o he 1997 Asian financial crisis. Sock reurn volailiy was found decreasing bu no ye calmed down o he level before he crisis by March He also poined ou ha he Asian financial crisis has he grea persisence effec on volailiy. Furhermore, a sudy on volailiy of sock reurns generaed from Jakara Sock Exchange Index during he Asian crisis was carried ou by Leeves (2007). The findings indenified asymmery impacs from condiional volailiy shocks indicaing negaive shocks led o greaer volailiy compare o posiive shocks during crisis. The asymmeric response sared o rise a he end of 1997 and hen declined during Though, he Indonesian sock marke urned o behave more symmerically o shocks owards he end of A he same ime, he long run persisence effecs declined. Karunanayake e al (2010) in addiion disclosed ha boh 1997 Asian financial crisis and 2008 financial crisis had no impac on mean sock reurns for U.S. bu posiive influence was deeced on sock reurn volailiy due o he crises. Besides, U.S. sock reurns demonsraed high volailiy persisence effec. Apar from he above, Kapopoulos and Siokis (2005) found evidence ha new generaion of crashes (crashes of lae 90s and early 00) exhibied sronger overreacion-overcorrecing paern compared o old generaion of crashes (crashes of lae 80s and early 90s). In oher word, he sock marke need longer urbulen period for a convergence o a new equilibrium seady sae afer old crashes. 2.2 Sock Marke Analysis and Global Even Despie of sock marke crash, sudies of sock marke had been done wihin he period covering major even especially he Sepember 11 h 2001 erroris aack. Charles and Darne (2006) validaed ha he erroris aacks a U.S. o be a permanen and emporary negaive shocks for mos of he inernaional sock markes. Ineresingly, he impac was more severe on oher world sock markes insead of he U.S. sock markes even hough he aacks were argeed direcly a U.S. Besides, hey also showed ha macroeconomics news announcemen from U.S. can generae large shocks o U.S. and European sock markes. On op of ha, Nikkinen e al (2008) compared he behavior of sock marke reurns and volailiy in 6 differen regions before and afer he erroris aacks a Sepember 11 h, They concluded ha here was significan increase in volailiy across regions afer he aacks while he sock reurns declined shorly afer he aacks bu rebounded over longer periods in 3 o 6 monhs ime. 2.3 Subprime Credi Crisis and is Conagion Effecs A research done by Dooley and Huchison (2009) explained ha here was a long genle decline in U.S. equiies a he sar of Subprime crisis in mid of 2007 hrough Sepember The sock marke indices in boh U.S. marke and emerging 104

4 markes exhibied a dramaic decline in Sepember wih exreme volailiy afer ha. The reason given was ha he news announcemens such as Lehman bankrupcy led o a massive impac on he equiies markes and ransmission of such impac o emerging markes was discovered. This finding is consisen wih he resuls by Celikkol e al (2010) regarding he increase of volailiy of Turkey price index (ISE- 100) afer he bankrupcy of Lehman Brohers. Besides, similar examinaion was employed by Longsaff (2010) exposed ha here was financial conagion spillover across o oher financial markes as he Subprime crisis developed. Anoher recen sudy was found by Ramlall (2010) regarding he influence of Subprime crisis on volailiy clusering and leverage effecs in major inernaional sock markes. Evidences revealed ha volailiy clusering had increase afer he Subprime crisis. The repor also suppored findings by Dooley and Huchison (2009) ha here was ransmission of Subprime crisis o oher emerging sock markes. A he same ime, he leverage effec in pos crisis was higher compare o pre crisis period in mos of he inernaional sock markes sudied. The volailiy due o negaive shocks was found o be more pronounced in pos crisis han pre crisis while he volailiy due o posiive shocks showed reverse paern. On op of he lieraures, Engle and Ng (1993) suggesed ha negaive shocks led o greaer impac on volailiy compared o posiive shocks when hey measured he effecs of news on volailiy esimaion using Japanese sock reurns from 1980 o In ha sense, his paper is moivaed o exend he sudies by invesigaing he impac of Subprime credi crisis on sock marke reurns and volailiy. To my knowledge, here are ye few sudies done o relae hese wo issues compared o he number of researches done on oher sock marke crashes. 3. The Mehodology and Model As he main objecive of his paper is o sudy he impac of Subprime crisis on U.S. sock marke reurn and volailiy, hence daily daa of S&P 100 sock indexes for he period of May 2006 o December 2009 were sudied. Invesigaion on S&P 100 sock indexes ha includes 100 leading U.S. socks allows his paper o generae a beer undersanding based on larges and mos esablished companies in U.S. The full period of sudy is hen divided ino 3 differen sages o allow behavior of sock marke reurns and volailiy o be invesigaed in each sage or sub-period of Subprime crisis. Referring o he facs and figures provided by Chrisie (2006), Isidore (2008), Dooley and Huchison (2009), he following sub-periods are developed: Early sage of he Subprime crisis (May 2006 o mid-dec 2007), Recession of U.S. (Mid-Dec 2007 o mid-sep 2008) and Period afer he Bankrupcy of Lehman Brohers (mid-sep 2008 o Dec 2009). 105

5 Figure 1: Sock Marke Price during Jan 2005 o December Model Specificaion Two main laen variables for his sudy are he sock marke reurn and volailiy. The sock marke reurns for each sub-period are calculaed as follows: R log P log P 1 where R is he daily reurn of S&P 100 indexes a day, P is he closing price a day and P -1 as he closing price a day 1. On he oher hand, volailiy is an unobservable variable. Hence, i is common pracice in lieraures (example see Perron and Qu 2008) o make use of square reurns or absolue reurns as proxy for volailiy. However, Andersen and Bollerslev (1998) commened ha squared reurns are poor esimaors of day-by-day movemens in volailiy and i is also an exremely noisy esimaor. In ha sense, he proxy for volailiy adoped in his paper follows he sep by Nikkinen e al (2008) which is he esimaed volailiy from he observed reurn series. As a preliminary analysis, uni roo es was employed on he observed reurn series for each sub-period using Augmened Dickey-Fuller (ADF) es and Phillips-Perron (PP) es by including he drif and consan erms. I was found ha all 4 series of reurns are saionary. In oher words, he reurn series have no uni roos. Hence, he level form of reurns will be used for furher esimaion hroughou he analysis. Nex, model specificaion of ARMA process hrough correlogram for he mean equaion was carried ou. All 4 series of reurns were found o fi AR(1) process well wih he minimum AIC value compared o oher ARMA process. Diagnosic check hrough various ess like inveribiliy es, auocorrelaion es (referring o Q- saisics for he correlogram of residuals) and normaliy es (Jarque-Bera es) were applied o verify he required assumpions. The resuls demonsraed ha 106

6 AR(1) process for all series do no violae any of he inveribiliy and nonauocorrelaion assumpion, however he normaliy assumpion is no saisfied. Hence, all he series of reurn were modelled using firs-order auoregressive process wih he following mean equaion: R 1 2R 1. (1) Neverheless, GARCH model will be applicable only if ARCH effecs are found. Once ARCH effecs discovered, he basic GARCH (1,1) model is adoped for esimaing he condiional variance in his paper suppored by mos of he researches conduced for volailiy sudy (example see Holden e al 2005, Charles & Darne 2006, Nikkinen e al 2008, Ramlall 2010). Therefore, he variance equaion will be presened as: h h. (2) where h is he condiional variance wih he condiions of 1 I ~ G. E. D 0, h and all parameers of GARCH (1,1) process are resriced o be non-negaive meaning o say 0, 0 and 0 o ensure ha he condiional variance (h ) is posiive. 3.2 Hypohesis Tesing Firsly, his paper aimed o check wheher here is any volailiy clusering deeced in differen ime periods. The signal of volailiy clusering can be esed on he ARCH coefficien in he condiional variance equaion (Eq. 2) which is capured by wih he following null hypohesis saemen: H 0 : i 0 where i represen he period/subperiods of sudy On he oher hand, his paper aemped o invesigae he persisence effecs of shocks on he volailiy of U.S. sock reurns wheher o be permanen or only ransiory. This objecive can be done by esing if he persisence measure is differen from uniy, i.e. H 0 : ( ) i 1 indicaing permanen effecs where i represen he period/sub-periods of sudy In wha follows, hypohesis esing on mean will be carried ou o examine wheher here is any significan change on he sock reurn and volailiy across he subperiods. Non-parameric es on mean via Mann-Whiney es will be performed for boh series if violaion of normaliy assumpion is found. The null hypohesis is presened as: 107

7 H 0 : i j where i and j represen he period/subperiods of sudy 4. The Findings Volailiy clusering in sock reurns implies ha large (small) changes in prices are followed by large (small) changes of eiher sign. The value of for he full period (0.086) presened in Table 1 showed significan volailiy clusering in whole. Ineresingly, he ARCH effec urned o be significan only in Sub-period III (0.0382) when we sudy on 3 differen sub-periods. Hence, here was only evidence of volailiy clusering in sock reurns afer he bankrupcy of Lehman Brohers. However, he ARCH parameer in full period or Sub-period III is less han uniy hus he shocks were no explosive. Moreover, he persisence measure in Table 1 provided informaion abou he impac of shocks on volailiy. I was found ha all he periods are insignificanly differen from uniy excep for Sub-period III. Therefore, evidence shows he persisence of shocks (Subprime crisis) o volailiy is permanen for boh Subperiod I and II implying ha he condiional variances are non-saionary. In conras, he shock of bankrupcy of Lehman Brohers due o Subprime crisis had only ransiory impac on he volailiy. In oher words, we perceived ha he persisence of shocks o volailiy decay afer he bankrupcy of Lehman Brohers a a slow rae as he persisence measure is very close o uniy. A he mean ime, hird research quesion o be answered in his aricle is abou he behavior of U.S. sock marke reurn and volailiy across he period of sudy a differen sages or sub-periods. Table 2.1 and 2.2 presened he summary saisics for boh reurn and volailiy series for each sages. On average, U.S. sock marke facing negaive reurns wih he onse of Subprime crisis hroughou he whole period of sudy referring o Table 2.1. The mean reurns declined from he early sage of Subprime crisis (Sub-period I) owards he recession period (Sub-period II) bu a small rebound was deeced afer he bankrupcy of Lehman Brohers. However, validaion of he changes on mean reurn was examined and he resul was disclosed in Table 3. The oucome indicaed ha he changes in mean reurns were no significan. In conrary, he volailiy of U.S. sock reurns rise wih he occurrence of Subprime crisis and coninuous incremen across sub-periods was discovered (refer he mean values in Table 2.2). Resuls from Table 3 in advance confirmed significan increase on volailiy while he magniude of increase in volailiy seems o be higher afer he bankrupcy of Lehman Brohers. 108

8 Table 1: AR(1)-GARCH(1,1) Resuls Mean Equaion Variance Equaion Persisence Value Period 1 2 Full * (3.643) * (-2.759) 7.86E-07 ** (1.753) * (4.111) * (49.102) (0.013) I * (3.682) (-2.457) 1.08E-06 (1.275) (2.381) * (25.570) (0.212) II (-0.629) (-1.439) 2.27E-05 (0.572) (0.334) * (3.474) (0.322) III (1.950) (-1.568) 5.56E-07 (0.524) * (2.023) * (51.155) *** (2.923) Noe: * and ** denoe significance a 5% and 10% level respecively wih z-saisics in he parenheses. *** implies significanly differen from uniy a 10% level wih Chi-Square saisics from Wald coefficien es in he parenhesis. Table 2.1: Descripive Saisics of Reurns Full Sub-period I Sub-period II Sub-period III Mean Median Sd. Dev Skewness Kurosis Jarque-Bera Saisics * * * Noe: * implies significanly deviaed from Normaliy assumpion a 5% level for JB es Table 2.2: Descripive Saisics of Volailiy Full Sub-period I Sub-period II Sub-period III Mean E Median E Sd. Dev E E Skewness Kurosis Jarque-Bera Saisics * * * * Noe: * implies significanly deviaed from Normaliy assumpion a 5% level for JB es 109

9 Table 3: Comparison of Reurns and Volailiy beween Sub-periods Comparison in Reurn Volailiy a. Mean Change Beween Sub-period I and II (II) (I) b. Mean Change Beween Sub-period II and III (III) (II) E E-04 * 5.738E E-04 * Noe: Mann-Whiney Tes (Non-parameric es) was conduced in order o es for he significan changes in he mean values of reurns and volailiy of U.S. marke across sub-periods. * denoe significance a 5% level 5. Summary and Conclusions As global evens and sock marke crash or crisis are an imporan elemen ha can have grea impac on he sock marke, hence i is worhwhile for an exend analysis of sock markes during Subprime credi crisis. This paper is moivaed o carry ou similar sudy by Dooley & Huchison (2009) and Ramlall (2010) o furher invesigae he behavior of U.S. sock marke reurns and volailiy since he Subprime crisis. Iniially, his paper found ha he volailiy ends o cluser for he whole period of sudy which is consisen wih he resuls by Ramlall (2010) of volailiy clusering afer Subprime crisis. However, his paper only deeced significan ARCH effecs afer he bankrupcy of Lehman Brohers in conras wih he early sage of Subprime crisis and recession period. Though, he persisence of shocks (Subprime crisis) on volailiy in he sudy before he bankrupcy was permanen while he opposie is appeared a is afermah. Therefore, we can perceive ha he shock caused by he bankrupcy has only ransiory impac on he volailiy of sock marke reurns and i decay in ime wih a very slow rae. Besides, U.S. sock marke revealed significan increases in volailiy hroughou he ime periods bu no on sock reurns. This is suppored by he previous lieraures like Chakrabari & Roll (2002), Law (2006), Leeves (2007) and Karunanayake e al (2010) due o crisis in general while on he oher hand consisen wih he findings of Dooley & Huchison (2009) and Celikkol e al (2010) in paricular due o he bankrupcy of Lehman Brohers. In conclusion, he news abou bankrupcy of Lehman Brohers following he Subprime crisis urned ou o have exacerbaed impac on sock marke volailiy bu no on he sock reurns in general. Anyhow, his aricle is limied wih volailiy modelling assuming symmeric effec. Hence, fuure research may furher he sudy by incorporaing wih asymmeric effec on volailiy esimaion due o sock marke crash. 110

10 References Andersen, TG & Bollerslev, T 1998, Answering he Skepics: Yes, Sandard Volailiy Models Do Provide Accurae Forecass, Inernaional Economic Review, vol. 39, no. 4, pp Celikkol, H, Akkoc, S & Akarim, YD The Impac of Bankrupcy of Lehman Brohers on he Volailiy Srucure of ISE-100 price Index, Journal of Money, Invesmen and Banking, Issue 18, <hp:// Chakrabari, R & Roll, R 2002, Eas Asia and Europe during he 1997 Asian Collapse: A Clinical Sudy of a Financial Crisis, Journal of Financial Markes, vol. 5, no. 1, pp Charles, A & Darne, O 2006, Large Shocks and he Sepember 11 h Terroris Aacks on Inernaional Sock Markes, Economic Modelling, vol. 23, no. 4, pp Choudhry, T 1996, Sock Marke Volailiy and he Crash of 1987: Evidence from Six Emerging Markes, Journal of Inernaional Money and Finance, vol. 15, no. 6, pp Chrisie, L 2006, Real Esae Cools Down, CNN Money, 16 May, viewed 14 April 2010, <hp://money.cnn.com>. Dooley, M & Huchison, M 2009, Transmission of he U.S. Subprime Crisis o Emerging Markes: Evidence on he Decoupling-Recoupling Hypohesis, Journal of Inernaional Money and Finance, vol. 28, no. 8, pp Engle, RF & Ng, VK 1993, Measuring and Tesing he Impac of News on Volailiy, The Journal of Finance, vol. 48, no. 5, pp Holden, K, Thompson, J & Ruangri, Y 2005, The Asian Crisis and Calendar Effecs on Sock Reurns in Thailand, European Journal of Operaional Research, vol. 163, no. 1, pp Isidore, C 2008, I s Official: Recession since Dec. 07, CNN Money, 1 Dec, reviewed 14 April 2010, <hp://money.cnn.com>. Kapopoulos, P & Siokis, F 2005, Sock Marke Crashes and Dynamics of Afershocks, Economics Leers, vol. 89, no. 1, pp Karunanayake, I, Valadkhani, A & O Brien, M 2010, Financial Crises and Inernaional Sock Marke Volailiy Transmission, Ausralian Economic Papers, vol. 49, no. 3, pp Law, SH 2006, Has Sock Marke Volailiy in he Kuala Lumpur Sock Exchange Reurned o Pre-Asian Financial Crisis Levels?, ASEAN Economic Bullein, vol.23, no. 2, pp Leeves, G 2007, Asymmeric Volailiy of Sock Reurns during he Asian Crisis: Evidence from Indonesia, Inernaional Review of Economics and Finance, vol. 16, no. 2, pp Longsaff, FA 2010, The Subprime Credi Crisis and Conagion in Financial Markes, Journal of Financial Economics, vol. 97, no. 3, pp Nikkinen, J, Omran, MM, Sahlsrom, P & Aijo, J 2008, Sock Reurns and Volailiy following he Sepember 11 Aacks: Evidence from 53 Equiy Markes, Inernaional Review of Financial Analysis, vol. 17, no. 1, pp Perron, P & Qu, Z 2008, Long-Memory and Level Shifs in he Volailiy of Sock Marke Reurn Indices, Available via hp://people.bu.edu/perron/papers/lmshifs.pdf. Ramlall, I 2010, Has he US Subprime Crisis Accenuaed Volailiy Clusering and Leverage Effecs in Major Inernaional Sock Markes?, Inernaional Research Journal of Finance and Economics, Issue 39, pp